New south wales

Appendix A – Setting and applying harvest thresholds

Setting thresholds for proportional threshold harvest strategies

Harvest strategies that use thresholds will not necessarily result in substantially lower yields. Research on proportional threshold harvesting (Lande et. al. 1997) indicated that average yield may even be increased if thresholds are set optimally. However, a drawback of threshold harvesting is that it may increase variance in annual yield since there may be some years when no harvesting is allowed if the population remains below the lowest threshold. Nevertheless, proportional threshold harvesting has been shown to be superior, in terms of reducing depletion and extinction while maintaining yield, to other harvesting strategies including proportional harvesting.

Threshold abundance levels can be set in a number of ways. Using a time series of abundance data, the threshold can be set at the minimum observed abundance (Deroba & Bence 2008). A potential disadvantage of this method is that the time series needs to be sufficiently long to be representative of the conditions (environmental and anthropogenic) that influence a population’s abundance, and so establish a reliable threshold. For example, if a rare event caused abundance to fall to a historically low level that is unlikely to occur again, the threshold might be set too low. Furthermore, if abundance falls below the threshold, which can happen even in the absence of harvesting, should the threshold be adjusted to the new low abundance or not? The somewhat arbitrary nature of the threshold can make management actions unclear when abundance falls below the threshold.

Alternatively, the threshold can be based on statistical properties of a time series of the population’s abundance. For example, a time series of abundance estimates can be plotted as a histogram (Figure 7). In this example, the distribution of abundance follows an approximately lognormal distribution with a mean of 15.2 kangaroos per square kilometre and a standard deviation of 5.8 kangaroos per square kilometre. In the long term, kangaroo density is expected to follow a lognormal distribution. This distribution can also be represented using z-scores. The z-score transformation quantifies the variables in terms of standard deviations from the mean. The z-score transformation also standardises the variables so that the mean of the distribution is zero and the standard deviation is one. The area under the curve between two z-scores represents the probability that an element of the distribution is the specified number of standard deviations from the mean (Figure 8). In terms of setting harvesting thresholds, a threshold set at a z-score of -1.5 would represent the lowest 6.7 per cent of the distribution, while a z-score of two represents the lowest 2.3 per cent of the distribution.

The advantage of this method of setting the threshold over a more arbitrary method is that the threshold is unlikely to be biased by a single low abundance. Additionally, as more survey data are added to the time series of abundance for a population, the estimates of the population’s mean and standard deviation become more robust.

Applying this method of setting thresholds to red kangaroos in harvest zone 2 (Figure 9) indicates an initial threshold of 7.8 red kangaroos per square kilometre and a lower threshold of 6.4 red kangaroos per square kilometre. If the annual aerial survey indicates that the population of red kangaroos is below 7.8 kangaroos per square kilometre, the annual quota is reduced from 17 to 10 per cent of the estimated population size. If the survey indicates that the population abundance of red kangaroos is below 6.4 kangaroos per square kilometre, then all harvesting in the zone will cease until at least the next survey when the annual harvest quota is reappraised. Thus, thresholds allow the population to fluctuate within its normal range in abundance, but prevent harvest mortality from depleting the population when it is at low abundance.

Density is estimated by aerial survey and the frequency of estimated densities is converted to probability densities. The distribution of kangaroo densities is approximately lognormal.

The mean of the distribution is zero and the standard deviation is one. Areas under the distribution represent probabilities. The orange shaded region represents the probability that a sample is between 1.5 and two standard deviations below the mean (and represents 4.4 per cent of the area). The red shaded region represents the probability that a sample is more than two standard deviations below the mean (and represents 2.3 per cent of the area).

The red line represents a normal probability distribution of the observed data with a mean of 15.2 kangaroos per square kilometre and a standard deviation of 5.8 kangaroos per square kilometre. The upper range of the orange region (7.8 kangaroos per square kilometre) represents the threshold within which harvest rate is reduced from 17 per cent to 10 percent. This lower rate is maintained unless density falls below 6.4 kangaroos per square kilometre, at which point harvesting ceases (red region). The thresholds were calculated after log transforming the data.

The following section shows how model simulations can be used to examine the relative effects of different thresholds applied to harvesting a theoretical population of red kangaroos.

Reducing the risk of overharvesting: an example using red kangaroos

Briefly, changes in kangaroo numbers are modelled as a function of pasture biomass which, in turn, is determined by recent rainfall, past pasture biomass and the density of kangaroos (and livestock) consuming the pasture. Harvesting obviously reduces kangaroo numbers, but the reduced density results in higher pasture biomass and therefore higher rates of increase of kangaroos. This improvement in environmental conditions for a population, which without harvesting has no long-term trend, is a basic requirement for the sustainability of a harvest. The population can be simulated 10,000 times over a 20 year period. Each run is different as, every three months, rainfall is drawn from a probability distribution using the average and standard deviation for rainfall in western NSW and thus reflects the uncertain food supply in this arid environment. Population size is also estimated with uncertainty by aerial surveys, and so this too was drawn from a probability distribution using the average and standard deviation associated with aerial surveys (Pople 2008). The population was harvested at an annual rate of 15 per cent or less if it was below a particular threshold.

Extinction is highly unlikely for this simulated population unless there is some combination of low numbers, catastrophic weather and unsustainable harvesting (i.e. much greater than 15 per cent). A more useful measure of threshold performance is the probability of the population dropping to a relatively low density. This can be calculated as the proportion of the 10,000 simulation runs where the population falls below particular densities. Thresholds can be expressed in terms of standard deviations (SDs) below long-term average density for a kangaroo management zone. That way, the aim of the threshold harvest strategy is to keep the harvested population above historically low density.

The effect of reducing harvest rate at varying SDs below the long-term average density for the simulated kangaroo population is shown in Figure 7. Reducing the threshold not surprisingly reduces the probability of very low density, but the decline in probability from no threshold (15 per cent harvest) to no harvest is smooth. There is therefore no obvious optimum with the choice being somewhat arbitrary. Notably, even an unharvested population has some chance of declining to very low density.

Standard deviation (over time) was calculated from a lognormal distribution. Mean population size was about eight kangaroos per square kilometre. Density was about four kangaroos per square kilometre at two standard deviations below the mean.

Other factors that could be considered in setting thresholds is the time spent below some relatively low density (e.g. Figure 8), and the long-term average and variability in harvest offtake (including years with zero offtake) (Pople 2003). For these simulations, average harvest offtake was similar among the thresholds shown in Figures 7 and 8, but variability in the annual harvest increased slightly as the threshold was reduced.



Figure 11. Simulated population as described for Figure 10.

Density was about two kangaroos per square kilometre at 4.5 standard deviations below the mean.

References

Deroba JJ & Bence JR (2008) A review of harvest policies: Understanding relative performance of control rules. Fisheries Research 94:210-223.

Engen S, Lande R & Saether B-E (1997) Harvesting strategies for fluctuationg populations based on uncertain population estimates. Journal of Theoretical Biology 186:201-212.

Hacker RB, McLeod SR, Druhan J, Tenhumberg B and Pradhan U (2004) 'Kangaroo management options in the Murray Darling Basin.' (Murray-Darling Basin Commission: Canberra).

Lande R, Engen S & Saether B-E (1995) Optimal harvesting of fluctuating populations with a risk of extinction. American Naturalist 145:728-745.

Lande R, Saether B-E & Engen S (1997) Threshold harvesting for sustainability of fluctuating resources. Ecology 78:1341-1350.

Milner-Gulland EJ, Shea K, Possingham H, Coulson T and Wilcox C (2001) Competing harvesting strategies in a simulated population under uncertainty. Animal Conservation 4: 157-167.

Pople A (2003) 'Harvest management of kangaroos during drought.' Unpublished report to New South Wales National Parks and Wildlife Service, Dubbo, NSW.

Pople AR (2008) Frequency and precision of aerial surveys for kangaroo management. Wildlife Research 35: 340-348.

Pople AR & Grigg G (1999) Commercial harvesting of kangaroos in Australia. At www.environment.gov.au/biodiversity/trade-use/wild-harvest/kangaroo/harvesting/index.html